Data-driven approaches for discovering perturbed interaction interfaces in cancer
Abstract: A major challenge in cancer genomics is the comprehensive identification of genes with functional roles in cancer ("drivers"). This is a difficult task as there is substantial mutational heterogeneity across tumors. Further, each tumor typically exhibits numerous mutations, whereas only a few are functionally relevant for tumor initiation or progression. Previous methods based on mutation recurrence analyses reveal common driver mutations across cohorts of patients, but inherently miss rare yet genuine driver mutations. Here, we introduce a method to identify genes with critical somatic mutations that preferentially alter their interaction interfaces, a pattern that has been observed across many known cancer drivers (e.g., TP53, IDH1, KRAS). We first develop an approach that combines large-scale sequence, domain and structure information to provide a robust, accurate, and biologically-relevant assessment of per-position ligand-binding potential across protein sequences. This enables us to pinpoint sites involved in interactions in 60% of human genes, representing the most comprehensive resource of this type to date. We next develop a novel analytical framework to integrate these domain binding potentials with additional sources of data to uncover genes whose interaction interfaces are significantly perturbed in tumors. Our method recapitulates known oncogenic and tumor suppressor cancer driver genes, and discovers novel, relatively rarely-mutated genes with likely roles in cancer. Finally, our interaction-based method can highlight perturbed molecular mechanisms stemming from individual cancer mutations, thereby enabling valuable insights that may help guide personalized cancer treatments.
Speaker Bio: Shilpa N. Kobren is a Ph.D. candidate at Princeton University in the Dept. of Computer Science and Lewis-Sigler Institute for Integrative Genomics. Her research is in computational and systems biology, where she focuses on detecting and interpreting protein interaction and cellular network perturbations within and across organisms in Dr. Mona Singh's lab. Before starting graduate school, Shilpa received her B.Sc. in Computer Science and Biology from Tufts University, where she worked on improving methods for protein structural alignments.